What happens to innovation after M&A waves?
Goodbye Wyeth, Schering-Plough, Genentech.
The wave of Pharma acquisitiveness has arisen anew after a short period of dormancy. It seems that every time this happens pundits, myself included, take a position on either side of the mega-merger debate Resolved: “Pharma mega mergers hinder pharmaceutical innovation and thus rob long-term investors of shareholder value”. Personally, I’ve always argued in the affirmative.
The dearth of quality academic research on the subject has perpetuated the debate. The biggest drawback to even quality work on this subject, like that of Patricia Danzon et al, is that the time horizon in such studies has been too short (usually 1 to 3 years) to capture the effects of M&A on long-cycle R&D outcomes and long-term equity investment. Generally speaking, though, academic studies suggest a modest negative effect of large mergers/acquisitions on R&D performance and a modest positive effect of R&D alliances. Recently, a couple of academic studies have been published that add more substance to this debate either by following firms that merge or acquire for longer periods of time post-event or by measuring more dimensions of R&D performance, thus offering a more complete picture of merger effects. I’ll discuss them both here and see whether the debate is tilted to one side by either or both of them.
I’ll begin with a book chapter by Grabowski and Kyle published last year that contains some new, albeit preliminary, research on the effects of mergers and acquisitions on drug-therapy development phase-transitions (i.e. project level), using a database (IMS’ R&D Focus) of 4500+ firms conducting R&D between 1990 and 2007 (the observation period includes mergers between 1985 and 2006). The researchers examined firms of all sizes, but since I’m primarily interested in large firms, I’m only going to discuss their findings for large and very large firms by their categorization; such firms have 20-50 and 50+ active R&D projects in the database per year, respectively. I’m also only going to report their findings for transitions from Phase 2 to 3 and beyond Phase 3, since public data on earlier work tends to be unreliable.
For large and very large firms, merging firms tended to have a higher fraction of projects advance from Phase 2 to Phase 3 within a 5-year window than non-merging firms. The fraction advancing for merging firms was roughly 24% for each size, whereas merging firms advanced ~33% of projects. With the Phase 3 to Launch transition, large non-merging firms appeared to under-perform their very large counterparts (49% vs. 57% advancing), and for both size classes, merging firms outperformed their non-merging counterpart: For large firms 49% vs. 65% advanced (non-merger vs. merger), and for very large firms 57% vs. 64% advanced. The probability of a project advancing to the next stage within five years was estimated by regression, controlling for firm size and year and merger was found to confer a higher probability of advancement.
This study has some major limitations, and Prof. Kyle tells me that she is in the process of expanding this line of research, which I hope will address some them. One of my biggest concerns is that it appears that a firm experiencing a merger at anytime during 1985-2006 is considered post-merger in the analyses, even if the project being examined transitioned in the pre-merger period, which eliminates any possibility of distinguishing cause from effect, i.e. do stronger R&D firms merge or do mergers result in stronger firms. But, even if the analysis had been more sophisticated, the question simply cannot be answered definitively by this type of study. Furthermore, Phase transition probabilities alone are hardly an adequate measure of R&D performance, even considering the all-important transition from Phase3 to Launch. What would be needed is one or more measures of the importance of the innovation advanced, either in terms of its utility (i.e. societal benefit) or its commercial success (i.e. investor benefit).
Let’s move on to the next study, from Prof. Carmine Ornaghi at the U. of Southampton, published this year in the International Journal of Industrial Organization. A working paper of this study from 2006 is also available. Ornaghi relies on public data from multiple sources covering the years 1988 through 2004 and focuses just on the largest firms (marketcap > $1B). From financial sources, Ornaghi examines revenues from approved drugs, total R&D expenditures, and marketcapitalization, using inflation-adjusted dollars. From a patent database, he culls the number of new patents, their technological classes (to infer technological relatedness), and the number of citations each patent received (to infer a patent’s overall importance). ATC codes were used to infer product relatedness. A total of 27 merger events were analysed.
Overall, mergers reduced research inputs (R&D spend total or as a percentage of revenues) and outputs (patents, patent citations), although the decline in R&D productivity (change in patents over R&D intensity) was not statistically significant. Overall return to shareholders during the three years following a merger was also reduced, although not significantly (p=0.06). An ex-ante analysis using a propensity score showed that firms with drugs approaching patent expiration and without new launched drugs are more likely to merge, confirming earlier work from Danzon. Matching firms using propensity scores and examining the period from four years pre-merger to three-years post-merger, Orrnaghi finds that the ex-ante characteristics of merging and non-merging firms are very similar, while innovation measures begin to differ following a merger. Using the matched-firm analysis, he does not find, however, that shareholder value is significantly reduced.
Looking at technology and product relatedness, the study suggests that product relatedness (measured by overlap in ATC codes) has a positive effect on R&D inputs and outputs and on shareholder value, whereas technology relatedness (overlap of patent classes) seems to have a deleterious effect. In other words, marketing synergies developed from similar portfolios of merging firms might help the combined firm in the marketplace, at least in the short run, but technological synergies seem harder to realize.
Like the earlier discussed work, the observational nature of this study rules out the possibility of definitively inferring causality. Nevertheless, the author uses some clever variable controls to gain insights not available from prior studies. Specifically, by matching firms on various observable characteristics present before a merger, he isolated-relatively speaking-the effects of merger per se on the post-merger performance. There is no way to account for nonobservable firm characteristics, of course, and I would have liked to have seen some additional control variables (e.g. firm experience across its TAs) but the approach used here is about as good as it gets for observational studies. The fact that mergers appear to degrade R&D intensity and productivity using this approach seems to tilt the debate in favor of the resolution.
So, what can we conclude from these additions to the Pharma M&A literature? The Phase transition findings of Grabowski and Kyle, if they hold, would seem to suggest an improvement in this measure of R&D output post-merger. But these preliminary results need to be fully fleshed out. I have questions about the methodology used, and Phase transition alone has serious shortcomings as a sole measure of R&D performance. In contrast, the study from Ornaghi is more complete. He provides some unique insights into the effects of merger on large-firm R&D productivity and shareholder value that argue against short-term benefits on either. The main drawback to this study is the follow-up period post-merger is too short–only three years. We need follow-up of five or more years to feel comfortable the findings are of high relevance to hughly innovative projects with long cycle times. I’d also like to see some additional control variables to further reduce selection bias, and additional outputs (such as Phase transition probability and surrogate measures of innovation importance) to better gauge the magnitude of merger effects. Overall, the weight of evidence still favors the Resolution in the affirmative, but I don’t think it would be fair to call the issue settled.
